The image generated by AIYXL using AI
A Near-Perfect Score Ends in a Public Confession
A candidate who achieved the highest overall score in South Korea's certified tax accountant examination has admitted that his exceptional performance was obtained through unauthorized assistance from artificial intelligence.
On November 15, the candidate, identified by the surname Han, acknowledged his actions during an appearance on a lawyer's YouTube channel.
According to Reuters and Yonhap News Agency, Han used electronic devices and generative AI during both stages of the examination, which took place in April and July 2026.
In the first stage, he answered 159 of 160 questions correctly, receiving 397.5 points out of a possible 400.
He subsequently achieved high marks in the second stage, ultimately ranking first overall.
The unusually strong results had attracted attention and suspicion in online communities.
As questions about the legitimacy of his performance intensified, Han publicly acknowledged the misconduct and submitted a written confession to the Seoul Metropolitan Police Agency.
The case has attracted national attention not simply because of the cheating itself, but because it illustrates how generative AI can be combined with concealed electronic devices to undermine established examination procedures.
How Generative AI Changed the Cheating Process
According to Han's account, he used a miniature camera to capture examination questions, transmitted the images to an AI system and checked the generated answers on a smartphone.
This approach differs from traditional cheating methods that rely on memorized answers or materials prepared in advance.
Generative AI systems can respond to previously unseen questions, potentially allowing users to obtain assistance without knowing the specific examination content beforehand.
The case illustrates three vulnerabilities in conventional examination environments: unauthorized capture of test materials, communication with an external information system and access to generated responses.
However, publicly available reporting has not established which AI model Han used or how accurately that system answered the examination questions.
It would therefore be misleading to attribute the outcome to any particular AI product.
South Korean police have confirmed receiving Han's statement and are investigating the case.
The relevant authorities have also announced plans to suspend the validity of his examination qualification.
Under applicable rules, confirmed cheating may result in invalidation of the examination and a five-year prohibition on taking the test.
Any potential criminal liability will depend on the investigation and subsequent legal proceedings.
Why Professional Certification Faces a New Challenge
Professional qualification examinations serve a purpose beyond ranking candidates.
They provide employers, clients, regulators and the public with evidence that individuals possess the knowledge and skills required for specialized responsibilities.
For professions involving taxation, accounting, engineering and other regulated activities, the reliability of certification can have significant consequences.
If candidates can secretly obtain AI-generated answers during examinations that prohibit external assistance, the relationship between examination results and actual professional competence becomes less reliable.
This creates challenges for examination authorities.
Traditional supervision generally relies on physical monitoring, restrictions on permitted materials and procedures designed to prevent communication with outside sources.
Miniaturized cameras, smartphones and AI-enabled devices can make unauthorized assistance harder to detect.
The South Korean case demonstrates why examination security may need to incorporate stronger device controls, improved monitoring procedures and more effective post-examination investigations.
It also raises questions about how examination authorities should respond when suspicious results emerge after certificates have already been issued.
Should Professional Examinations Ban AI or Redesign Assessment?
The broader implications extend beyond preventing cheating.
Artificial intelligence is increasingly used legitimately in professional workplaces for research, document analysis, calculations and preliminary decision support.
This creates an important distinction between assessing independent knowledge and evaluating the ability to work responsibly with AI tools.
For examinations intended to measure personal competence without external assistance, restrictions on AI use remain essential.
However, some professional assessments may eventually incorporate authorized AI tools while requiring candidates to verify outputs, explain their reasoning and identify errors.
Such approaches would evaluate a different set of skills.
They would not eliminate the need for independent knowledge, particularly in professions where incorrect decisions can create financial, legal or safety consequences.
A credible assessment system may therefore require multiple methods, including controlled examinations, practical exercises and evaluations of professional judgment.
AIYXL Analysis: When Answers Become Easier, Verifying Expertise Becomes Harder
The central lesson of this case is not that artificial intelligence has no place in education or professional practice.
Rather, it is that the widespread availability of powerful AI systems makes the verification of individual competence increasingly important.
Educational institutions must reconsider how they assess learning.
Professional certification organizations must strengthen the reliability of their examination processes.
Technology developers and policymakers may also need to consider how AI-enabled devices affect environments where independent performance is required.
AI can make information more accessible and help professionals work more efficiently.
But access to an answer is not equivalent to understanding it, validating it or accepting responsibility for its consequences.
In an AI-enabled economy, professional credibility may depend increasingly on the ability to demonstrate reliable judgment beyond the answers generated by machines.
Sources:
Reuters , November 15, 2026, Top scorer on South Korean tax exam confesses to cheating with AI.
Yonhap News Agency, November 15, 2026, Top scorer in tax accountant exam turns himself in over AI-assisted cheating.
Editorial note: The 159 correct answers refer specifically to the first stage of the examination, which contained 160 questions. The full qualification process included a second stage.
